Apple has acquired Spektral, a computer vision firm situated outside Denmark has experience in dealing with segmentation technology, for more than $30 million. Segmentation technology serves as an efficient technique used to “cut out” figures, particularly from backgrounds in digital videos and images.
This kind of technology can be utilized, for instance, in making more accurate and quicker cut-out images in augmented reality settings. It can also be used for extra standard applications such as school photos.
In fact, this serves as the initial market that was targeted by the startup back in 2015, even though it later appeared to change strategy in a bid to make deeper inroads into video and build up IP.
Rumors of the Apple and Spektral deal surfaced recently after Borsen, a Danish financial newspaper, revealed the details without Apple ’s confirmation.
However, Apple confirmed the details after we reached out by saying: “Apple buys smaller technology companies from time to time, and we generally do not discuss our purpose or plans.”
From our understanding, the acquisition took place a while back. In fact, this thought corresponds with Toke Jansen’s Linked profile. Jansen has been Spektral ’s co-founder but currently notes that as of Dec 2017, he has been managing Apple ’s computational imaging department.
Other individuals associated with the company including Henrik Paltoff, the other Spektral co-founder, are yet to update their profiles. As such, it is still uncertain how many more people have joined.
On the other and, Borsen reports that the acquisition includes Spektral’s engineers and was estimated to be around 200 million Danish kroner or nearly $31 million.
Spektral began life at CloudCutout and leveraged algorithms obtained from Jansen’s Ph.D. Initially, the startup advertised its product as a more efficient and cheaper “green screen” technology in a bid to get rid of primary images from their plain or standard-pattern backgrounds.
The early version of the product was created through training the system on more than 100,000 professional cutouts.
Segmentation could aid in adding live filters to human figures in a picture. Moreover, it can be effective in occluding augmented reality environments, especially behind figures in a bid to make digital AR content to seem like they are interacting with the position of human beings.
What’s more, Spektral ’s segmentation tech can run on mobile phones, making it a more efficient and quicker way of processing AR images on devices. One of its key selling aspects is how it deals with very fine elements like hair.
“To provide high-quality cutouts, the core of our engine exploits recent advances in spectral graph theory and neural networks. The computation of pixel transparencies (the alpha channel) for a single image involves solving multiple large-scale equation systems, as well as carrying out multiple feed-forward passes in our neural networks,” said the founders when they fundraised a seed round.
“We pose the problem of determining an alpha channel of an image as a machine learning task. Compared to usual chroma keying, this allows us to consider a much broader range of backgrounds since the model will learn, i.e., texture representations from existing training data.”